AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

An Intriguing Property of Geophysics Inversion

Yinan Feng, Yinpeng Chen, Shihang Feng, Peng Jin, Zicheng Liu, Youzuo Lin

ICML 2022 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 4 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace 1D sine encoder kernel with sin(n pi t)cos(n pi x) (ablation)InvLINT · MarmousiMAE: 136.67 -> 128.33Improved
replace 1D sine encoder kernel with sin(n pi t)cos(n pi x) (ablation)InvLINT · Kimberlina-LeakageMAE: 8.13 -> 8.88Hurt
replace Gaussian decoder kernel with sin(n pi x)sin(n pi z) (ablation)InvLINT · MarmousiMAE: 136.67 -> 177.36Hurt
use Gaussian with 1/3 sigma or sinc kernel instead of Gaussian decoder kernel (ablation)InvLINT · MarmousiMAE: 136.67 -> 138.19Hurt
replace transformer+linear decoder with Conv x2 + Linear (ablation)InvLINT · MarmousiMAE: 136.67 -> 140.72Hurt
replace transformer+linear decoder with Deconv x1 + Linear (ablation)InvLINT · MarmousiMAE: 136.67 -> 167.98Hurt
use deeper (Deconv + Conv) x5 decoder instead of 1 transformer layer (ablation)InvLINT · MarmousiMAE: 136.67 -> 126.59Improved
use deeper (Deconv + Conv) x5 decoder instead of 1 transformer layer (ablation)InvLINT · Kimberlina-LeakageMAE: 8.13 -> 6.86Improved
use unshared multiple linear layers instead of a shared linear layer (ablation)InvLINT · MarmousiMAE: 136.67 -> 138.82No clear effect
increase transformer depth from 1 to 3 layers (ablation)InvLINT · MarmousiMAE: 136.67 -> 132.19Improved
increase transformer depth from 1 to 3 layers (ablation)InvLINT · Kimberlina-LeakageMAE: 8.13 -> 8.09No clear effect
reduce kernel number N from 2048 to 1024 (ablation)InvLINT · Kimberlina-LeakageMAE: 8.13 -> 8.63Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy